Main

Over the past few decades, the African hominin fossil record predating 3.0 million years ago (Ma) has expanded to include several potential ancestors for later taxa4,5,6,7. Currently, systematic analyses place A. afarensis as the most likely candidate for the middle Pliocene stem taxon from which multiple later hominin genera (that is, Homo and Paranthropus), and possibly other species of Australopithecus, descended8,9,10. A. afarensis was geographically and temporally widespread, and fossil sites with its remains are known from Tanzania to northern Ethiopia and, potentially, Chad. However, temporally, A. afarensis is not known after 2.95 Ma (ref. 11).

Paranthropus and Homo are well-documented in the eastern African fossil record after 2.0 Ma, especially in the Omo-Turkana Basin and at northern Tanzanian localities12,13,14,15, but the hominin fossil record between the last appearance of A. afarensis (around 2.95 Ma) and 2.0 Ma is patchy. For example, no Paranthropus fossils have been recorded from the Afar Region to date, despite its presence in the Omo-Turkana Basin and at Nyayanga, Kenya2 at approximately 2.7 Ma and in the Upper Ndolanya Beds at Laetoli16 at about 2.66 Ma. Additionally, a Homo specimen at Ledi-Geraru, Ethiopia, extends the genus closer in time17 (2.78 Ma) to the last known appearance of A. afarensis1,11. However a simple cladogenic model of A. afarensis, or any other stem taxon, splitting into these daughter genera is complicated by the presence of A. garhi in the Afar3 at approximately 2.5 Ma.

The Ledi-Geraru Research Project (LGRP) area is located towards the northern extent of palaeoanthropological sites in the Afar Region, Ethiopia (Extended Data Fig. 1). New discoveries in Ledi-Geraru suggest that early Homo and Australopithecus were both present in the Afar Region before 2.5 Ma, just as early Homo and Paranthropus are sympatric in the Omo-Turkana Basin and sites to the south18,19 after about 2 Ma. Whether the apparent absence of Paranthropus from the Afar Region reflects the spotty nature of the fossil record or a biogeographical signal is yet to be determined. What is clear is that the Afar Region has currently yielded the only definitive evidence for Australopithecus in eastern Africa after 2.95 Ma—A. garhi and the newly discovered specimens from Ledi-Geraru.

Ledi-Geraru geologic context

The LGRP area contains fossiliferous sediments from the critical 3.0–2.0 Ma time period17. The fossil sites in the Lee Adoyta and Asboli regions of the LGRP area are located west of the Awash River in a region incised by the Mille and Geraru River drainages and their tributaries (Extended Data Fig. 1). The 3.0–2.5 Ma fossiliferous sediments are exposed in fault blocks bounded by post-depositional northeast–southwest and nothwest–southeast trending faults (Fig. 1); age control is provided by many dated and correlated tephra deposits and by magnetostratigraphy17. Detailed 40Ar/39Ar dating methods are described in Methods.

Fig. 1: Geologic context of the Lee Adoyta and Asboli regions.
figure 1

a, Geologic map of the Lee Adoyta basin. Interbedded tuffs are shown as thin coloured lines. The black line indicates the position of the A–A′ cross-section shown in d. b, Stratigraphic sections at hominin localities showing the stratigraphic level of hominin fossil discoveries (yellow bones), fossiliferous horizons (white bones) and marker beds (coloured tuffs). Fossil and stratigraphic section locations are shown in a,c,d. c, Geologic map of the Asboli region superimposed on a hillshade image. Outcrops of multiple tephra deposits are shown as red lines, and these include the Giddi Sands and Lee Adoyta Tuffs that also occur in the Lee Adoyta region. d, South–north geologic cross-section of the Lee Adoyta basin showing the locations of hominin fossil sites. Vertical scale is doubled relative to the horizontal (2×VE).

Full size image

The distribution and stratigraphy of fossiliferous units and tephra deposits provide context for the hominin fossils described here, which were discovered in sediments cropping out in the Lee Adoyta and Asboli regions (Fig. 1 and Extended Data Fig. 1). The Gurumaha sedimentary package is present in narrow fault-bounded exposures in the central Lee Adoyta basin and in drag-faulted blocks adjacent to basalt ridges bounding the basin to the east (Fig. 1a). The Gurumaha Tuff (GT), a light grey lapilli tuff dated to 2.782 ± 0.006 Ma (1σ; recalculated17,20), provides age control for the unit (Fig. 1b). Stratigraphically younger, the Lee Adoyta sedimentary package is widely exposed in the Lee Adoyta basin and is correlated to exposures in Asboli (Fig. 1). The Lee Adoyta Tuffs (LAT) comprise two thin, geochemically distinct tephras: a yellow altered basaltic ash which occurs approximately 10 cm above a white rhyolitic ash dated to 2.631 ± 0.011 Ma (1σ; recalculated17,20). These tuffs, and an underlying olive-green clay, provide an excellent marker sequence (Fig. 1b and Extended Data Fig. 2a). The Giddi Sands sedimentary package (Tgs; containing the Giddi Sands Tuff (GST)) and the Markaytoli sedimentary package (Tmk; containing the Markaytoli Tuff) crop out in the eastern Lee Adoyta basin (Fig. 1a). The base of the Giddi Sands sedimentary package is an erosional unconformity cutting into the Lee Adoyta sedimentary package in the Lee Adoyta Basin (Fig. 1b). The GST is a 3- to 8-cm-thick laminated, multicoloured (orange, yellow and white), crystal bentonite tuff. 40Ar/39Ar single-crystal incremental heating (SCIH) yielded a weighted-mean age of 2.593 ± 0.006 Ma (1σ) for the GST sampled in Asboli (Methods and Supplementary Figs. 1–3). In Asboli, the Giddi Sands sedimentary package containing the AS 100 fossil site unconformably overlies a sequence of mudstones containing the Asboli Tuffs, which in turn overlies the LAT exposed approximately 500 m to the south (Fig. 1c).

New Ledi-Geraru hominin specimens

Hominin dental specimens (Fig. 2) were recovered from the three sedimentary packages described above—the Gurumaha, Lee Adoyta, and Giddi Sands (Table 1). Comparative methods and sampling protocols are described in Methods.

Fig. 2: New hominin dentition from the LGRP.
figure 2

Right, from top: LD 302-23 P3, LD 750 P4, AS 100 M1 and AS 100 M2. Left, images show the LD 760 assemblage (top, from left: maxillary molar, I2, I1, maxillary canine; bottom, mandibular molars).

Full size image
Table 1 New Ledi-Geraru hominin dental specimens with geological contexts, ages and identifications
Full size table

Gurumaha hominin specimen

A mandibular right third premolar, LD 302-23 (Fig. 2 and Extended Data Fig. 3), was recovered approximately 22 m to the southwest and 7 m stratigraphically below the LD 350 early Homo mandible locality1, above the Gurumaha Tuff17 (Fig. 1b). An enamel fragment is missing from the lingual corner, but the crown is otherwise well preserved. The crown is shorter mesiodistally (MD) than buccolingually (BL) (11.5 mm BL × 10.5 mm MD; Extended Data Fig. 3) and the major crown axis is oriented buccolingually. A relatively small metaconid sits well mesial to the protoconid, creating a tiny, centrally placed anterior fovea that is bounded by a low, but continuous, mesial marginal ridge. Occlusal wear makes it difficult to determine if the metaconid would have been completely distinct from the protoconid at the outer enamel surface. The talonid is mesiodistally short and occupies a relatively small portion of the crown area in occlusal view.

LD 302-23 departs strongly in shape and occlusal form from P3s attributed to Australopithecus anamensis and Australopithecus deyiremeda, and unicuspid specimens of A. afarensis (for example, A.L. 128-23 and A.L. 288-1), in which the major axis of an asymmetric crown runs mesiobuccal to distolingual and in having a continuous mesial marginal ridge that creates a ‘closed’ anterior fovea7,21,22 (Extended Data Fig. 3a). Some A. afarensis specimens (such as A.L. 437-2) resemble the crown outline of LD 302-23 (Extended Data Fig. 3a); however, LD 302-23 also differs from those specimens in having a mesially shifted metaconid, which creates an acute angle between the transverse crest and mesial protoconid crest, and in having a highly reduced anterior fovea relative to crown size23. In crown shape index (MD/BL), LD 302-23 falls in the middle range of the A. afarensis distribution (Extended Data Fig. 3b,c). In area (BL × MD), LD 302-23 falls within the range of the A. afarensis size distribution, at the lower limit of the range of Paranthropus boisei, but within the range of values of Paranthropus robustus (Extended Data Fig. 3b,d). Unlike Paranthropus12, the talonid of LD 302-23 is not expansive. We attribute LD 302-23 to Homo for a number of reasons. Although the sample of P3s attributed to early Homo is morphologically diverse (for example, KNM-ER 5431, OMO 75-14, KNM-ER 1802 and OH 7), LD 302-23 is clearly derived relative to pre-3.0 Ma Australopithecus, lacks the suite of derived nonmetric features characteristic of Paranthropus, and is consistent with the size, crown morphology and mesiodistal compression of the LD 350-1 specimen (the LD 350-1 P3 crown is damaged mesiolingually, with only around 65% of the crown remaining), which was found in the same sedimentary package and shares numerous dental and mandibular apomorphies with Homo.

Lee Adoyta hominin specimens

An isolated P4, LD 750-115670 (Fig. 2, Extended Data Fig. 4 and Supplementary Fig. 4), was collected from locality LD 750 at the base of an approximately 8-m-thick fossiliferous exposure of mudstone and sandstone. The stratigraphic position of the locality lies between the 2.63 Ma LAT and the 2.59 Ma GST (Fig. 1b and Extended Data Fig. 2a). The LD 750 P4 is unworn, with salient mesial cusps (Extended Data Fig. 4a). The roots are broken beneath the crown, but the mesial portion preserves a maximum root height of about 2.0 mm. The tooth may not have been fully erupted, as there are no interproximal contact facets and the occlusal surface is unworn. The protoconid and metaconid are shifted relatively mesially on the crown and tightly compressed BL, giving the tooth a slightly ‘puffy’ appearance. The anterior fovea is distinct and deep, but relatively small (reflecting the mesial placement of the cusps), centrally placed and bounded by a narrow mesial marginal ridge. The talonid is relatively large, the distal marginal ridge is low and rounded, and a distinct, but low, distolingual cusp is present, giving the tooth an asymmetric occlusal profile.

With dimensions of 12.4 mm BL × 11.4 mm MD, the tooth area is in the upper size range of A. afarensis, exceeded in area only by A.L. 996-1 (Extended Data Fig. 4b). Morphologically, however, the LD 750 P4 departs from the typical Hadar A. afarensis appearance. For example, the lingual to mesial face transition is smooth and rounded, whereas in Hadar P4s there is a more distinct corner. The occlusal layout differs from that of Hadar P4s in that the metaconid and protoconid are placed more centrally in relation to buccolingual crown breadth and shifted mesially in relation to crown length. Thus, compared with Hadar A. afarensis teeth, the buccal face is more convex occlusocervically and the anterior fovea more confined. The index of protoconid–metaconid apex distance/BL breadth shows the central placement of the LD 750 mesial cusp apices in relation to crown breadth compared with A. afarensis (Extended Data Fig. 4c,d).

Not only is LD 750 large relative to most A. afarensis specimens, it also exceeds the area of all specimens attributed to A. anamensis and all but two specimens of Australopithecus africanus (StW 498 and StW 384). The LD 750 crown area falls within the size distribution of P. robustus and Paranthropus aethiopicus; however, it lacks any of the distinct ‘molarization’ seen in Paranthropus P4s (for example, expanded talonid and distal cusps, thick marginal ridges, symmetrically rounded occlusal profile caused by the filling of the crown distobuccally, and surface complexity in unworn specimens12). We note that the known specimen of A. garhi lacks any mandibular remains3, but LD 750 cannot be excluded from that species on the basis of size alone (see randomization analysis in Extended Data Fig. 4e). In occlusal form and crown outline, LD 750 does resemble OMO 75-14 (Extended Data Fig. 4a), which has been attributed to aff. Homo by Suwa et al.12 and Hlusko et al.24, and KNM-ER 5431, which is also attributed to Homo by Suwa et al.12. However, these attributions are provisional and based mostly on the nonmetric morphology of their associated molars (for example, presence of C7 on M1 and M2 of both specimens). In fact, P4 morphology poorly distinguishes between early Homo and Australopithecus12,24. Thus, the phenetic resemblance between OMO 75-14, KNM-ER 5431 and LD 750, which are all large but lack features typical of Paranthropus, could be most parsimoniously interpreted as symplesiomorphic retentions. Since LD 750-115670 lacks any clear Homo synapomorphies associated with other specimens of early Homo from Koobi Fora (such as talonid reduction in KNM-ER 99225), we provisionally assign it to aff. Australopithecus.

An associated set of five mandibular molars, a maxillary molar fragment, and three maxillary anterior teeth (Fig. 2 and Table 1) representing a single individual were collected at locality LD 760, a relatively flat-lying area dominated by fine-grained sand and silts (Supplementary Video 1). The hominin teeth were all found in close association, in a relatively dense cluster among scattered faunal elements (Extended Data Fig. 2b). The mandibular molars were found eroding from the surface and the maxillary dentition was sieved from the surface silt (approximately 5–10 cm of loose and eroded surface soil) downslope on a gentle gradient over a 7 × 4 m area to the northwest. The LD 760 collection area lies approximately 10 m below the 2.631 ± 0.011 Ma LAT17 mapped in hillslopes surrounding the site (Fig. 1a and Extended Data Fig. 2a,b). Mammalian fossils were also observed eroding from sand units around 3.5 m and 6.5 m above the LAT (Fig. 1b). However, on the basis of the slope and distance to these sand units, as well as the abundance and spread of other fossil specimens, the hominin teeth and other fossils on the LD 760 surface represent a distinct fossiliferous unit below the LAT.

The LD 760 mandibular molars are moderately worn, with dentine exposure visible at buccal cusp apices in the M1 (Fig. 3). They are relatively wide BL compared to their MD length (Extended Data Figs. 5 and 6) and do not taper strongly distally, which gives the crowns, especially M1 and M2, a squarish profile (Fig. 4). The buccal grooves are indistinct, and the buccal profile is convex in occlusal view, in part because a protostylid partially fills in the buccal groove. Further, the buccal crown face slopes occlusocervically, reflecting an internal placement of cusp apices.

Fig. 3: LD 760 molars compared to A. afarensis.
figure 3

Left molars from Ledi-Geraru specimen LD 760 (left) and Hadar specimen A.L. 400-1 (right). Measurements in mm of the LD 760 molars (BL × MD): LM1: 13.3 × 13.4, LM2: 14.5 × 14.6, LM3: 14.0 (estimated) × 15.7, RM1: 13.2 × 13.1, RM2: 14.8 × 15.2 (Supplementary Data 1). Specimens are oriented with their buccal surfaces to the left and mesial surfaces up.

Full size image
Fig. 4: Contour of LD 760 compared to A. afarensis.
figure 4

Detail of distal taper (orange lines) and bilobate contour (blue lines) seen in A. afarensis M1 specimen A.L. 400-1 (right) contrasted with the overall more equilateral occlusal profile of the LD 760 M1 (left). The M2s also show this pattern (Fig. 3). Specimens are oriented with their buccal surfaces to the left and mesial surfaces up.

Full size image

There is no hint of M3 reduction (as seen, for example, in some early Homo such as LD 350-1 and OMO-75-14); instead, the molars follow the plesiomorphic pattern of relative molar size, M1 < M2 < M3. The M1 and M2 are subequally square (Extended Data Figs. 5 and 6), unlike in early Homo which tend to have rectangular crowns that are mesiodistally elongated. Further, the teeth are relatively simple, with few accessory cusps; the M1s show no expression of a C7, unlike LD 350-1 and KNM-ER 1802, and the M2s also lack a C7, which is seen in many specimens attributed to early Homo (for example, OMO-75-14, KNM-ER 5431, KNM-ER 60000 and OH 7). The mesiodistal dimensions of the mandibular molars comfortably fit within the metric variation seen in individual A. afarensis molars (Extended Data Fig. 5), and the buccolingual breadths fall in the upper third of the A. afarensis distribution. However, the sum of the molar areas for the LD 760 individual (614.3 mm2) exceeds all but one (NFR-VP-1/29; Woranso-Mille, Afar Region, Ethiopia) of the 11 A. afarensis individuals that preserve all 3 molars; yet, among Paranthropus, only 1 individual (DNH 7) is smaller in M1–M3 area than LD 760 (Extended Data Fig. 5d). The LD 760 molars depart from the typical feature set of A. afarensis in three ways: (1) the M1s are buccolingually broad relative to their mesiodistsal length (A. afarensis M1s and M2s are buccolingually widest across the mesial cusps, the LD 760 M1 is equally wide at midcrown and the M2 is only slightly narrower at midcrown) (Extended Data Figs. 5 and 6); (2) the M1 and M2 do not taper distally to the extent seen in A. afarensis (Figs. 3 and 4); and (3) the LD 760 molars lack the distinct bilobate buccal contour seen in A. afarensis12 (Fig. 4).

A fragmentary maxillary molar, LD 760-115911, was recovered at the LD 760 locality (Extended Data Fig. 7a). Although the protocone is the only fully preserved cusp, the lingual groove is distinct occlusally and the mesial edge of the hypocone is present. On A. garhi upper molars, the lingual groove is indistinct, but the lingual margin of the crown begins to curve buccally distal to the lingual groove, reflecting the relatively small hypocone and buccolingually narrow distal crown (Extended Data Fig. 7a). This reduction in hypocone area produces a triangular form for the Bouri specimen’s protocone. The LD 760 upper molar fragment shows no similar curvature of the crown margin distal to the lingual groove. Instead, it appears to have had a fully expressed hypocone. However, in terms of absolute molar area (BL × MD), the ratio of LD 760 lower molars to the A. garhi upper molars does not falsify a single-species hypothesis using modern ape comparative data—see resampling analysis in Extended Data Fig. 6b.

A right maxillary canine, a complete left maxillary lateral incisor and a left fragmentary maxillary central incisor were also found at the LD 760 locality (Fig. 2 and Table 1). The canine (LD 760-115979) is well preserved, with all but the root apex present (Fig. 5). The crown has mesial and distal interproximal contact facets. A distal interproximal contact facet is also present on the LI2, which indicates bilateral absence of a canine/I2 diastema; a diastema is present on the BOU-VP-12/130 A. garhi maxilla and is variably present in A. afarensis. The canine crown is worn apically and along the mesial and distal crests. Apical wear makes it difficult to examine the relative placement of the crown shoulders, but the mesial shoulder appears to have been placed apically. Despite the wear, the canine is not small (Extended Data Fig. 8a–c), and the ratio of the canine to molar mesiodistal dimensions puts this specimen in the upper range of Australopithecus and well beyond the observed values for Paranthropus (Extended Data Fig. 8d).

Fig. 5: Comparative maxillary canine morphology.
figure 5

a, Lingual (left) and labial (right) views of the Ledi-Geraru LD 760-115979 canine (left) with Hadar A. afarensis specimens A.L. 763-1 (middle) and A.L. 333x-3 (right). Note that the LD 760 canine is a right canine, whereas the A. afarensis canines are from the left and are mirrored in these images. b–d, LD 760-115979 (b; shown in lingual view) contrasted with Hadar A. afarensis specimen A.L. 199-1 (c; right canine shown; distal to the upper right) and Bouri A. garhi specimen BOU-VP-12/130 (d; left canine, mirrored; distal is to the right). Note the simple mesial–distal chisel-like wear pattern on the LD 760 canine (b) in contrast to the complex multi-faceted wear pattern of A. afarensis (c) and the broad curved basin on the distal side of the A. garhi upper canine (d; this is seen on both left and right canines). Images are oriented differently to emphasize the distinctive relevant morphology. Photos of BOU-VP-12/130 A. garhi holotype fossil by T. White; use courtesy of the Middle Awash research project. Images in b–d are not to scale.

Full size image

The LD 760 canine departs subtly from the macrowear of A. afarensis (Fig. 5). For example, its wear pattern lacks the characteristic ‘j-pattern’ along the distal crest26; instead, apical wear dominates. In basal size (10.4 mm BL × 10.5 mm MD) the tooth resembles A. afarensis and is narrower than the labio-lingually broad canine of A. garhi (Extended Data Fig. 9). The LD 760 canine also departs morphologically from A. garhi. The A. garhi canine, uniquely among hominins, possesses a broad, shallow distal basin reminiscent of a talon (Fig. 5d); this smooth basin is contiguous with a wide wear furrow that characterizes the entire postcanine dental row (Extended Data Fig. 9a).

The LD 760 individual cannot be attributed to Paranthropus. It lacks the derived reduction of anterior tooth size, which is reflected both in the absolute sizes of the anterior teeth and their size relative to the postcanine dentition (Extended Data Fig. 8d). The mandibular molars do not express relatively large C6s12 and they show a buccolingual wear gradient distinct from the planar pattern typical of Paranthropus12. The LD 760 teeth also lack any derived Homo traits (such as steeply walled molars, narrow M1, C7 presence and M3 reduction). We regard the morphology of LD 760 as inconsistent with known eastern African Australopithecus species because: (1) the lack of any of the derived A. garhi traits, notably the reduced hypocone of the maxillary molars and highly distinctive upper canine ‘talon’ basin; and (2) a morphology derived relative to A. afarensis (that is, lack of bilobate mandibular molar walls and distal taper, and canine wear pattern). Because of the plesiomorphic resemblance to A. afarensis, we make a generic-level attribution and assign the LD 760 assemblage to Australopithecus sp. indet.

Giddi Sands hominin specimens

A partial left M1 crown with lingual cusps only partially preserved (AS 100-1a) and 2 fragments assembled into a complete left M2 crown (AS 100-1b) were discovered eroded from the approximately 1-m-thick Giddi Sands unit in the Asboli region immediately below the GST (Fig. 2, Table 1 and Supplementary Fig. 5). Subsequent screening produced no additional hominin material.

Wear on these teeth is slight, with crown apices salient and no dentine exposed. Interstitial wear is present and the interproximal contact facets are congruent, consistent with identification of AS 100-1a as an M1 and AS 100-1b as an M2 (Fig. 2 and Extended Data Fig. 7). The AS 100-1b M2 is rhomboidal in occlusal outline, with the cusps offset such that the buccal pair sits mesial to the lingual pair, the hypocone is relatively large and projects distolingually, and the lingual profile is straight. A. afarensis M2s range in shape from squarish (for example, A.L. 199-1, A.L. 486-1) to more rhomboidal (for example, A.L. 444-2, A.L. 200-1)27, but A. afarensis specimens invariably show more pronounced lingual occlusocervical sloping, indicating a more internally placed cuspal arrangement than is apparent for AS 100-1b. The lingual cusps of the AS 100 molars are relatively steep vertically and lack the ‘puffy’ contour seen in relatively unworn specimens of A. afarensis (for example, A.L. 200-1). The rhomboidal crown outline, relatively large and projecting hypocone, and cuspal placement of AS 100-1b are closely matched by A.L. 666-1, a Homo specimen from the Busidima Formation at Hadar28 from 2.3 Ma, and MLP-1549, a Homo specimen from Mille-Logya29, 2.5–2.4 Ma. This feature set strongly contrasts with the A. garhi M2 in which the hypocone is relatively small and placed transverse to the metacone, the lingual profile is strongly curved, and the cusps are more internally placed (Extended Data Fig. 7a). In size, the AS 100-1b M2 (14.2 mm BL × 13.9 mm MD) and Homo specimen A.L. 666-1 are nearly identical and both slightly exceed the Mille-Logya M2 in BL breadth; AS 100-1b is smaller in area than the M2 of A. garhi and eastern African Paranthropus (Extended Data Fig. 7b). We attribute AS 100-1 to Homo sp. indet. based on the presence of a derived M2 crown outline (that is, rhomboidal with projecting relatively large hypocone30), a feature shared with other specimens of early Homo.

Taxonomic and phylogenetic implications

The presence of both early Homo and Australopithecus at Ledi-Geraru has implications for hominin taxonomy and diversity in this region in the 3.0–2.0 million year interval. Despite the relative paucity of fossils discovered in this time interval, evidence for multiple non-robust lineages in eastern and southern Africa indicates that taxonomic diversity had already evolved by 2.5 Ma. Here we examine taxonomic and phylogenetic hypotheses for the newly discovered hominin specimens from Ledi-Geraru.

First, although the Asboli sample contains only two molars (Table 1), and they predate Homo specimens A.L. 666-1 and MLP-1549 by more than 150,000 years, we regard the most parsimonious hypothesis to be that these are members of the same species of Homo30. The existence of Homo at Ledi-Geraru by around 2.78 Ma was previously established by the LD 350-1 mandible1; the new dental material from the Asboli area, as well as the LD 302-23 premolar described here from the Gurumaha sedimentary package, provide additional evidence of Homo prior to 2.5 Ma.

Second, LD 750 and LD 760 are separated by 24 m of strata and straddle the LAT. We provisionally regard all teeth recovered from the Lee Adoyta sedimentary package as representatives of a single species of Australopithecus, partially premised on our expectation of the low likelihood of two Australopithecus species existing in such close geographic and temporal proximity. The following are potential phylogenetic scenarios for the LD 750 premolar and the LD 760 teeth:

  1. (1)

    The Ledi-Geraru australopith teeth represent a late-surviving population of A. afarensis. The Lee Adoyta sample is approximately 350,000 years younger than the last appearance of A. afarensis, which is documented by specimens from the Kada Hadar 2 Submember at Hadar11. Contradicting this hypothesis, the mandibular M1s and M2s are not bilobate and are squarer in crown shape than the majority of A. afarensis; the P4 and molar sizes fall just at the upper limit of size variation in that species; and the pattern of maxillary canine wear in the LD 760 canine is not seen in A. afarensis. Accordingly, if these teeth do represent A. afarensis, they may capture further evolutionary change within the lineage. The resemblances between the Lee Adoyta Australopithecus specimens and A. afarensis are symplesiomorphic. In addition to A. garhi, this is the only other record of the persistence of this genus in eastern Africa after 2.95 Ma, although it has approximate contemporaneity with A. africanus found in South Africa.

  2. (2)

    The Lee Adoyta Australopithecus teeth represent a population ancestral to later Paranthropus, which is currently undocumented in the Afar Region; this inference could be supported by the large size of the postcanine teeth. However, the appearance of Homo by 2.78 Ma implies a divergence between that clade and Paranthropus by at least that date, and Paranthropus is present as early as 2.66 Ma in the Upper Ndolanya Beds at Laetoli, Tanzania16 and around 2.7 Ma at Nyayanga, Kenya2. Moreover, the Lee Adoyta teeth share no synapomorphies with Paranthropus (such as presence of C6 on the molars, reduced anterior tooth size or planar wear pattern) and megadonty is common in early hominin lineages, including A. garhi, A. africanus and early Homo3. Given these constraints, it is unlikely that the Lee Adoyta Australopithecus is ancestral to Paranthropus.

  3. (3)

    The Lee Adoyta Australopithecus teeth are early representatives of A. garhi. Accepting this hypothesis would require that A. garhi includes very different canine and maxillary molar forms and postcanine wear patterns on their upper and lower dentitions. This alternative is difficult to evaluate as there are only a few overlapping anatomical elements to compare between the current A. garhi fossil material and the Lee Adoyta Australopithecus sample. However, our assessment is that the lack of similar forms on the few shared elements suggests the Lee Adoyta australopith is not A. garhi.

  4. (4)

    The Lee Adoyta teeth are a previously unknown species of Australopithecus from the early Pleistocene. This hypothesis is the only alternative that offers no contradictions with the data presented in the previous three hypotheses.

The Ledi-Geraru fossils described here demonstrate that there were at least 3 lineages in the Afar Region between 3.0 and 2.5 Ma: Homo, A. garhi and the newly identified Australopithecus from Ledi-Geraru. The diversity of hominins in the interval around 2.5 Ma demonstrates the ways in which evolution was experimenting with the overall hominin pattern; in addition to these Afar region hominins, A. africanus is present in South Africa and Paranthropus is found in Kenya, Tanzania and southern Ethiopia by this time. The apparent absence of Australopithecus after approximately 2.0 Ma in eastern Africa means that only two hominin genera remained, Homo and Paranthropus, and these were well differentiated in terms of their dietary ecology31. Moreover, the discoveries of Australopithecus and Homo at Ledi-Geraru between 2.78 Ma and 2.59 Ma reveal that a drier, more open habitat was not uniquely associated with the appearance of Homo17,32, raising intriguing questions about niche breadth and landscape use among early hominins—including why Australopithecus was able to survive in the Afar Region alongside Homo until at least 2.5 Ma, and whether Paranthropus was competitively excluded from the Afar Region by late-surviving Australopithecus filling a similar niche.

Methods

40Ar/39Ar dating

Individual phenocrysts of K-feldspar (sanidine to anorthoclase, ~400–1,000 μm) from tuff sample LG-1048-1 of the Giddi Sands Tuff were analysed at the Berkeley Geochronology Center by the SCIH 40Ar/39Ar dating technique17. The feldspar concentrate was irradiated in the Cd-lined CLICIT position of the Oregon State University TRIGA reactor for four hours (BGC irradiation no. 461). Sanidine phenocrysts from the Alder Creek Rhyolite of California were used as the neutron-fluence monitor mineral, with an orbitally referenced age20 of 1.1848 ± 0.0006 Ma. Reactor-induced isotopic production ratios for this irradiation was: (36Ar/37Ar)Ca = 2.65 ± 0.02 × 10−4, (38Ar/37Ar)Ca = 1.96 ± 0.08 × 10−5, (39Ar/37Ar)Ca = 6.95 ± 0.09 × 10−4, (37Ar/39Ar)K = 2.24 ± 0.16 × 10−4, (38Ar/39Ar)K = 1.220 ± 0.003 × 10−2, (40Ar/39Ar)K = 2.5 ± 0.9 × 10−4. Atmospheric 40Ar/36Ar = 298.56 ± 0.31 (ref. 33) and decay constants follow ref. 34.

In total, 26 individual grains were analysed separately by the SCIH approach. One grain was rejected after the initial step as an old xenocryst (27361-08, ~100 Ma), while all others yielded geologically reasonable eruption ages and were carried to conclusion (Supplementary Table 1). All completed experiments demonstrated apparent-age plateaus, typically across the entirety of the 39Ar release spectrum (Supplementary Fig. 1), indicating an absence of complexity to the internal argon systematics, and that these feldspars show no evidence of alteration or inherited argon. A probability density plot of the plateau ages is shown in Supplementary Fig. 2, revealing that the age distribution is symmetrical and gaussian-like. The weighted-mean age of the accepted plateaus is 2.593 ± 0.006 Ma (1σ analytical error). Isochrons (36Ar/40Ar versus 39Ar/40Ar) calculated from the plateau steps yield an indistinguishable weighted-mean age of 2.597 ± 0.006 Ma (Supplementary Fig. 3) but, given the high proportion of radiogenic 40Ar of total 40Ar (typically >90%) in these crystals, we prefer the straightforward plateau weighted mean as the reference age for this sample (2.593 ± 0.006 Ma).

Dental analysis

Hominin specimens discovered in 2015 and 2018 from the LGRP area are described and evaluated taxonomically, using comparative fossil material from eastern and southern Africa including A. anamensis, A. afarensis, A. africanus, A. deyiremeda, A. garhi, Paranthropus aethiopicus, P. boisei, P. robustus, Homo habilis, Homo rudolfensis, and Homo sp. (see Extended Data Table 1 for sites, localities and references for these taxa). Resampling data from Pan troglodytes and Gorilla gorilla metrics are also used to evaluate the comparative relative variation within and between samples. Bootstrapping (resampling with replacement) was performed in R v. 4.4.135. The analyses were conducted using code written by the authors that utilizes the sample function. For each analysis, distributions of values for extant taxa were derived from 10,000 bootstraps. The ratio of the Ledi-Geraru and A. garhi specimens were compared to the bootstrapped distributions for chimpanzees and gorillas; all statistical tests were two-tailed using α = 0.05. All photographed material (including comparative specimens) were examined directly by the authors. Quantitative data came from a variety of sources, including the authors’ and published measurements; both quantitative and qualitative comparative analyses were performed. See Extended Data Table 1 for references and Supplementary Data 1 for all hominin dental metrics.

Survey and collection for each locality

LD 302-23

This mandibular premolar was found 6 February 2015 by K.G.T. during palaeontological survey ~22 m to the southwest and ~7 m stratigraphically below the LD 350 early Homo mandible locality1, above the 2.782 ± 0.006 Ma Gurumaha Tuff17 (Fig. 1b). The fossil was recorded in the LGRP database with GPS coordinates. Numerous non-hominin fossils were also recovered in the area.

LD 750-115670

This mandibular premolar was found by O.A.O. during palaeontological survey on 14 February 2018 (Supplementary Fig. 4). It was located ~9 m stratigraphically above the 2.63 Ma Lee Adoyta Tuff, on the eastern side of the Lee Adoyta Basin (Extended Data Fig. 2a) The fossils were recorded in the database with GPS coordinates and an in situ photograph. The sequence includes fine sandy silts to very fine sand, tuffaceous silts and brown mudstones.

LD 760-115329, LD 760-115685, LD 760-115533, LD 760-115316, LD 760-115323, LD 760-115979, LD 760-115934, LD 760-115884 and LD 760-115911

The first molar, LD 760-115329, an M1, was found by O.A.O. on 14 February 2018, followed by the discovery of four more lower molars by A.L.R. and O.A.O. within ~30 min. These teeth were located ~10 m stratigraphically below the 2.63 Ma Lee Adoyta Tuff on the eastern side of the Lee Adoyta Basin (Extended Data Fig. 2). All fossils surrounding the hominin teeth were flagged and collected during surveying for additional hominin material (Supplementary Video 1). The fossils were recorded in the database with GPS coordinates and in situ photographs. The area was subsequently (19–25 February 2018) screened with 2 mm screen mesh and the maxillary dentition was discovered in the surrounding 4 × 7 m in the upper 5–7 cm of loose surface material. Expanded sieving in 2020 did not produce additional fossils. Fossil survey, collection, and screening are shown in Supplementary Video 1. The sediments below the Lee Adoyta Tuff at the site include fine sands overlain by brown mudstones.

AS 100-1a and AS 100-1b

On 23 February 2015, one partial hominin molar, AS 100-1a (LM1), was found by B.M.A. and a tooth fragment, AS 100-1b, was subsequently found by E.M.L. during palaeontological survey (Supplementary Fig. 5). Further exploration of the small area revealed a second fragment, found by B.M.A., which fit onto AS 100-1b (completing the LM2). All teeth were found within 0.5 m2. All fossils were recorded in the database with GPS coordinates. The area was subsequently screened with 2 mm screen mesh without recovering more fossils. The sequence includes ~1 m thick sand (Giddi Sands) and thin pebble conglomerate units immediately below the Giddi Sands Tuff (Fig. 1).

For all localities, geological data were collected using StraboSpot 2.12.18, Avenza 5.4 and GaiaGPS 2025.2 and integrated in ArcGIS Pro 3.2.2.

Ethics and inclusion

The Ledi-Geraru Research Project has long partnered with local Afar team members to successfully conduct palaeoanthropological research in the Afar Region, Ethiopia. Three Ledi-Geraru Afar team members whose contributions and discoveries were critical for this study are recognized as co-authors. The Ledi-Geraru Research Project is committed to formally acknowledging the contributions of all our collaborators and team members in this and future work. This acknowledgement is one step in our long-term commitment to building research capacity and access for the local Afar community.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.